Obscure bug in distributed density-matrix calcTotalProb
TysonRayJones opened this issue · 1 comments
TysonRayJones commented
There is a bug in calcTotalProb
when a density matrix is distributed in such a way that each node contains fewer than one-column (e.g. a half column). For example, when a 3
qubit density-matrix (64 elements between 8 columns) is distributed between 16 nodes (4 elements per node, i.e. 1/2 column).
The bug is a simple algorithmic error in the counting of local diagonal elements, and always occurs for the aforementioned distirbuted pattern. Luckily, this is the same obscure scenario as reported in #322. Unluckily, it does not cause an error, but simply returns an incorrect result!
TysonRayJones commented
patched in #328